A high-speed, high resolution (i.e. approximately 1 mil and finer) 3-D laser scanning system for inspecting miniature objects such as circuit board components, solder, leads and pins, wires, machine tool inserts, etc., can greatly improve the capabilities of machine vision systems. In fact, most problems in vision are 3-D in nature and two-dimensional problems are rarely found.
Several methods have been used to acquire 3-D data: time of flight, phase detection, autofocus, passive stereo, texture gradients, or triangulation. The latter approach is well suited for high resolution imaging and is perhaps the most well known technique.
In the general scanning triangulation method a laser beam is scanned across the object to be inspected with a deflector and the diffusely scattered light is collected and imaged onto a position sensitive detector The scanner can be a rotating polygon, galvanometer, resonant scanner, holographic deflector, or acousto-optic deflector Likewise, the position sensitive detector can be a linear or area array sensor, a lateral effect photodiode, a bi-cell, or an electro-optic position sensing device. Sometimes, a pair of position detectors are used to reduce shadowing. With linear arrays or area cameras there is severe trade off between shadows, light sensitivity and field of view.
For obtaining very high speed and low light sensitivity, the position sensing system described in the above-noted patent application is preferred However, if it is not required to detect very low light levels, lateral effect photodiodes can be used at data rates up to about 1 MHz and are inexpensive, commercially available devices.
Often triangulation-based methods and systems have used the concept of "structural light". As described in U.S. Pat. No. 4,105,925 such a method involves projecting a line or multiple lines onto the surface of the object to be inspected and detecting the displacement of the projected line (or multiple lines) with a video camera. Such systems are now available off-the-shelf and are relatively inexpensive.
The primary disadvantages of such a system are the very low speeds (typically 10,000 points/second) and, in the case of multiple projected lines in a single image, ambiguous interpretations of the data result from overlap of adjacent stripes and multiple scattered light between stripes. Both disadvantages can be overcome by replacing (1) the line projector with a flying spot scanner and (2) the video camera with one of several types of position sensitive detectors, as illustrated in U.S. Pat. No. 4,375,921.
Conventional triangulation based scanners or structured light systems often utilize conventional imaging lenses (i.e., reduction lenses, 35 mm lenses, or cylinder lenses designed for long line detectors) to deliver light to large area position sensitive detectors such as area sensor, linear arrays or large area position sensitive detectors The large area detectors have several limitations: low speed due to large detector capacitance, high dark currents, and a much higher noise floor than what is found with small area devices.
For example, a 20 mm.times.20 mm P-I-N lateral photodiode (equivalent to the approximate area of a typical 1" video camera tube) has a capacitance of several hundred picofarads and a dark current of several microamps. On the other hand, a 2 mm.times.2 mm device will have capacitance of about 5 pf and a dark current of about 50 nanoamps. Both the speed and noise performance of the smaller detectors can be orders of magnitude better than the performance achievable with large area devices. The improvement in speed is directly proportional to the reduction in capacitance and the improvement in signal-to-noise is at least as large as the square root of the reduction in capacitance.
With typical triangulation-based images it is difficult to deliver light to a small area device without decreasing the field of view (and consequently the inspection speed). Furthermore, if the field of view is increased the height resolution is necessarily decreased in conventional triangulation based imagers. Also, if a spherical reduction lens is used to deliver light to the detector (with the necessary proportional decrease in resolution) the light gathering capability of the system is reduced in proportion to the area. These are severe limitations and impose undesirable trade-offs which limit the system performance.
A "synchronized scanning" approach can be used to overcome this problem as described in U.S. Pat. No. 4,553,844 to Nakagawa et al. This scanning approach is commonly implemented with polygonal or galvanometer driven mirrors. However, this approach requires that the sensor head contain moving parts in the form of a rotating mirror (for example, in the Fournier plane or telecentric stop) or a pair of mirrors. In effect, a second mirror is used to follow the spot which is scanning by means of the first mirror. These moving parts are often not desirable, particularly if the sensor is to be subjected to the type of acceleration found with x-y tables and robotic arms in industrial environments.
A dilemma exists with conventional triangulation imagers: it is desirable to use a small detector but unless moving parts are included the field of view becomes too small, the resolution too coarse, and the light gathering capability poor. Even if the coarse resolution is tolerable, the loss of light gathering capability also further reduces the system signal-to-noise ratio The signal-to-noise ratio is not good in the first place (particularly at high speeds) because of the use of the large area detector thereby compounding the problem.
Many other prior U.S. patent describe various methods for the acquisition of 3-D data by means of triangulation. For example, the U.S. Pat. No. 4,188,544 to Chasson describes a structured light method in which a beam expander and cylinder lens is used to project a line of light onto an object. The line of light is sensed with an imaging lens and video camera. The position of each point is determined with a peak detection algorithm. The measurement rate is slow due to the readout of the video camera. Multiple lines of light alleviate this problem to some extent.
In the U.S. Pat. No. 4,201,475 to Bodlaj, an object is scanned in a position sensing dimension and the time displacement is detected by a single photodetector having a very narrow field of view. The speed of the system is limited by the retrace time of the scanning device at each measurement point. This method is relatively slow especially for the requirements of small part inspection at quasi-video rates (i.e. MHz).
In the U.S. Pat. No. 4,645,917 to Penny, a swept aperture profiler is described. It too measures a time displacement for determining position. A galvanometer driven mirror is used to scan a line of data (i.e. x, y coordinates). An acousto-optic deflector is used to scan the position sensing dimension and the instant at which the light is received by the photodetection device indicates depth. The use of the A-O deflector for the z dimension scanning represents an improvement over the previous technology. Also, the use of a photomultiplier as a detection device allows for a much improved dynamic range.
The U.S. Pat. No. 4,355,904 to Balusubramanian, describes a triangulation-based method which incorporates a position sensing device in the form of a variable density filter together with a system for sweeping the laser beam and controlling the position of the measurement probe. The tolerance on the density of typical variable filters, whether fabricated with a metallic coating on glass or with photographic film plate, is typically +5% at any single point.
The U.S. Pat. No. 4,589,773 to Satoshi Ido, et al., describes a position sensing method and system for inspection of wafers which utilizes a commercially available position detector. A reduction lens is used to focus the light into a small spot on the surface of the object with a 10X reduction. A magnification lens is used in the receiver (10X) to deliver light to a detector. The triangulation angle is 45 degrees with the receiver and detector at complementary angles (90 degrees). This is fine for wafer inspection. However, the method is deficient for several other types of inspection tasks because (1) unacceptable shadows and occlusion effects would occur for tall objects; (2) the field of view of the probe is very small; (3) a reduction of the angle to 15 degrees (to reduce shadows) would degrade the height sensitivity significantly; and (4) the detector area is relatively large which limits speed and the signal to noise ratio as the speed of the system is increased.
The U.S. Pat. No. 4,472,056 to Nakagawa et al., describes a method which involves projection of a line of light and the use of a rectangular CCD as the position sensor. This represents a significant improvement in speed over the method described in the above noted U.S. patent to Chasson and is good for inspection of parts with a relatively limited height range (i.e. 16 levels). Logic and hardware is included for peak detection which can be related to the depth of the object.
In the U.S. Pat. No. 4,650,333 to Crabb et al., a method of structured light projection is described which is somewhat complementary to the method described in the Nakagawa patent noted immediately above. A stripe of light produced with a cylindrical lens is swept across the object with an acousto-optic deflector in such a way that a single CCD line array can be used. This is a less expensive way of implementing the structured light method which does not require a custom CCD. Again, the speed and stray light rejection capabilities of the probe are limited which restrict it to depth measurement of objects (like traces) which are not very tall. Nevertheless, the method is suited to the inspection task of trace height measurement.
The U.S. Pat. No. 4,593,967 to Haugen assigned to Honeywell describes a triangulation-based scanning system utilizing a holographic deflection device to reduce the size and weight of the scanning system and a digital mask for detection of position. The digital mask is in the form of binary grey code and requires a detector for each bit (i.e. 8 detectors for an 8 bit code). A single cylinder lens is used in the receiver to convert a spot of light into a thin line which must be sharply focused onto a series of photodetectors. In other words, the spot is converted into a line to deliver the light to the series of long thin detectors. Spatial averaging is not performed in the system nor is the centroid of the light spot determined.
U.S. Pat. No. 4,634,879 discloses the use of optical triangulation for determining the profile of a surface utilizing a prism and two photomultiplier tubes in a flying spot camera system. These are arranged in a "bi-cell" configuration. The bicell, however, does not compute the centroid of the received light spot and is therefore sensitive to the distribution of intensity within the received light spot. As an anti-noise feature, amplitude modulation is impressed upon the laser beam and a filter network is used to filter photomultiplier response so as to exclude response to background optical noise.